Our Ph.D. student Juan Francisco Marin Vega has combined his love for technology with his interest in image editing to create a solution that is aiming to enableing image transformations without human supervision.
The project named Image enhancement with single/multiple exposures (AIERE) strives for automation of color and brightness image transformation with deep neural networks. The first step is to reduce the time required to perform these tasks with the eventual possibility to completely automate them.
The scale of the project requires great technical skills and the most advanced techniques from the computer vision field. Because high resolution images can be difficult to process there are some challenges that require a combination of techniques to solve.
He is working with Python and makes use of open-source frameworks such as NumPy, PyTorch, fast.ai and PyTorch-Lightning (along with many more). To accelerate the computations, these models perform and require access to GPUs.
With the success of this project we can empower the image editing business with no limits. We can automate tasks that are applied to hundreds of images every day, freeing human editing time to more challenging and compelling tasks.
After a few months of trial and evaluations with QA, some candidate models have been created and are being evaluated in a safe-production environment. So far there have been promising results in the area of single-exposure image enhancement.